<p>This study explores the effectiveness of AI-assisted formative assessment in a university-level C++ programming course. The approach integrates generative AI (GenAI) across a range of assessment tasks, from in-class quizzes to ill-defined problem-solving reports. The AI-generated evaluations serve not only as personalized feedback for students but also as real-time guidance for instructors to dynamically adapt their instructional practices. To examine the impact of this approach, a comparative study is conducted between an experimental group, which receives continuous AI-generated formative feedback, and a control group, which receives only feedback on their reports. Results indicate that AI-assisted evaluation enhances feedback efficiency and provides instructors with valuable diagnostic insights, such as aggregated error patterns visualized through word clouds. Additionally, student feedback suggests a generally positive perception of the AI-supported teaching approach. Moreover, human regrading procedure illustrates that while GenAI demonstrates strong potential for enabling personalized formative assessment, challenges remain in ensuring scoring consistency and precision. Implement details are discussed and the prompts provided to the ChatGPT-4o are also included in this work. We hope this research serves as a pedagogical case for enhancing the effectiveness of GenAI-assisted teaching.</p>